Re: ext2 + -osync: not as easy as it seems

[Eric Sandeen <sandeen@redhat.com>]

Jens Axboe wrote:
> On Wed, Jan 14 2009, Jan Kara wrote:
>>> I'm not sure what you mean; if the barrier operation isn't flushing
>>> all of the caches all the way out to the iron oxide, it's not going to
>>> be working properly no matter where it is being called, whether it's
>>> in ext4_sync_file() or in jbd2's journal_submit_commit_record().
>>   Well, I thought that a barrier, as an abstraction, only guarantees that
>> any IO which happened before the barrier hits the iron before any IO which
>> has been submitted after a barrier. This is actually enough for a
>> journalling to work correctly but it's not enough for fsync() guarantees.
>> But I might be wrong...
> 
> It also guarentees that when you get a completion for that barrier
> write, it's on safe storage. Think of it as a flush-write-flush
> operation, in the presence of write back caching.

(sorry for chiming in so late)

Jens, isn't this just the way it's implemented today?  At some point
couldn't a barrier bio simply be a reordering barrier that the storage
can use when destaging the write cache, rather than the heavy-handed
flush-write-flush we have today?

I guess it's a question of the intended semantics of a barrier bio, vs.
today's implementation based on current hardware functionality...

-Eric